Hi there!

I am Xuanlei Zhao, a second-year PhD student in Computer Science at National University of Singapore advised by Yang You, where I also completed my master’s studies. I obtained my bachelor’s degree in CS & EE from Huazhong University of Science and Technology. I currently intern at Adobe Research with Yan Kang and Yuanjun Xiong. Previously, I collaborated at Pika with Chenlin Meng and interned at Colossal-AI with Jiarui Fang.

My current research mainly focuses on efficient AI, including:

• Efficient diffusion and autoregressive models, e.g., for video generation.
• Efficient machine learning system, with parallelism and low-level optimization.
• Efficient foundation model customization, with parameter generation.
• Co-optimization of algorithm and infrastructure.

🔥 News

• 2025.07: Join Adobe Research as research intern in Seattle.
• 2025.05: DSP accepted by ICML 2025!
• 2025.01: PAB accepted by ICLR 2025 and integrated into Diffusers!
• 2024.03: Release VideoSys (OpenDiT), an efficient training and inference system for video models.
• 2024.02: HeteGen accepted by MLSys 2024!
• 2024.01: AutoChunk accepted by ICLR 2024!
• 2024.01: Start my PhD journey!

📝 Selected Publications (all)

🎬 Efficient Video Generation

• ICLR 2025 Real-Time Video Generation with Pyramid Attention Broadcast
  [paper] [code] [blog]
  
  Xuanlei Zhao^*, Xiaolong Jin^*, Kai Wang^*†, Yang You^†
• ICML 2025 DSP: Dynamic Sequence Parallelism for Multi-Dimensional Transformers
  [paper] [code]
  Xuanlei Zhao, Shenggan Cheng, Chang Chen, Zangwei Zheng, Ziming Liu, Zheming Yang, Yang You
• Training Variable Sequences with Data-Centric Parallel
  [code] [blog]
  Geng Zhang^*, Xuanlei Zhao^*, Kai Wang^†, Yang You^†

⚙️ Efficient System Optimization

• ICLR 2024 AutoChunk: Automated Activation Chunk for Memory-Efficient Long Sequence Inference
  [paper] [code]
  Xuanlei Zhao, Shenggan Cheng, Guangyang Lu, Jiarui Fang, Haotian Zhou, Bin Jia, Ziming Liu, Yang You
• MLSys 2024 HeteGen: Heterogeneous Parallel Inference for Large Language Models on Resource-Constrained Devices
  [paper]
  Xuanlei Zhao^*, Bin Jia^*, Haotian Zhou^*, Ziming Liu, Shenggan Cheng, Yang You
• PPoPP 2024 FastFold: Optimizing AlphaFold Training and Inference on GPU Clusters
  [paper] [code]
  
  Shenggan Cheng, Xuanlei Zhao, Guangyang Lu, Jiarui Fang, Tian Zheng, Ruidong Wu, Xiwen Zhang, Jian Peng, Yang You

🕹️ Efficient Model Customization

• Drag-and-Drop LLMs: Zero-Shot Prompt-to-Weights
  [paper] [code] [blog]
  
  Zhiyuan Liang^*, Dongwen Tang, Yuhao Zhou, Xuanlei Zhao, Mingjia Shi, Wangbo Zhao, Zekai Li, Peihao Wang, Konstantin Schürholt, Damian Borth, Michael M. Bronstein, Yang You, Zhangyang Wang^*, Kai Wang^*

💡 Open-Source Projects

• VideoSys (Project Lead): An Easy and Efficient System for Video Generation
• Colossal-AI (Top Contributor): Making large AI models cheaper, faster and more accessible
• FastFold (Top Contributor): Optimizing AlphaFold Training and Inference on GPU Clusters

💻 Internships

• 2025.07 - now, Adobe Research with Yan Kang and Yuanjun Xiong.
• 2024 - 2024, Pika with Chenlin Meng.
• 2022.07 - 2023.12, Colossal-AI with Jiarui Fang and Shenggui Li.

📖 Educations

• 2024.01 - now, PhD in Computer Science, National University of Singapore
• 2022.08 - 2023.12, Master in Computer Science, National University of Singapore
• 2018.09 - 2022.06, Bachelor in Computer Science & Electrical Information, Huazhong University of Science and Technology

💬 Invited Talks

• 2024.07, Real-Time Video Generation with Pyramid Attention Broadcast, Ventures [video]
• 2024.07, Speedup for Video Generation, Bytedance internal talk